AI Engineer | MS Machine Learning @ Carnegie Mellon | Production LLM agents (PydanticAI, MCP, RAG)
I design and ship full-stack AI products: typed LLM agents, MCP tool servers, RAG pipelines, and observability-first backends. AI Engineer at Enterprise Solutions; previously BNY Mellon and CMU research.
| Project | Stack | Description |
|---|---|---|
| Robinhood AI Portfolio Copilot (repo) | FastAPI, PydanticAI, MCP, Next.js | Live at myportfoliocopilot.com — Robinhood sync, macro pulse, risk detection, streaming AI chat |
| AI Generated Podcast | Python, LLMs, TTS | End-to-end news podcast pipeline - sample on Spotify |
| LLM Stack | LangChain, OpenAI, RAG | Collection of LLM apps: RAG chatbots, PDF Q&A, summarization, prompt engineering |
| Portfolio Website | Next.js, TypeScript | Dual-mode portfolio site (recruiter + creative modes) |
| Area | Tools |
|---|---|
| LLM Agents | PydanticAI, tool calling, sub-agents, memory, SSE streaming |
| Tool Protocols | Model Context Protocol (MCP), read-only tool isolation |
| Backend | FastAPI, async SQLAlchemy, PostgreSQL, Redis, JWT |
| Observability | Pydantic Logfire, OpenTelemetry |
| Frontend | Next.js, React, TypeScript, TanStack Query, shadcn/ui |
| ML / Research | PyTorch, diffusion, contrastive learning, computer vision |
|
|
|
|
|
|
|
|
|
|
|
|
- Agent architectures with MCP and provider-agnostic LLM layers
- Production guardrails: rate limits, PII scrubbing, read-only tool boundaries
- Macro-aware portfolio analytics and explainable risk scoring